1 |
On 12/07/2015 20:51, Alex Thorne wrote: |
2 |
> I'm afraid I won't be testing it any time soon -- I don't have any |
3 |
> drives to pair at the moment. As for your comments about dmraid being |
4 |
> 'fake', I'm a little confused. From what you say it sounds like this is |
5 |
> the hardware RAID that comes with many motherboards. Why is hardware |
6 |
> RAID undesirable over software RAID? Presumably the mdraid software |
7 |
> option has an associated performance hit? Is it just that opaque |
8 |
> proprietary firmware can't be trusted with the important task of looking |
9 |
> after our data? |
10 |
|
11 |
It's because the "RAID" abilities built into most motherboards are |
12 |
really shitty. Very little, if any, optimization going on, no real |
13 |
intelligence, and the whole thing just looks and feels like it's no more |
14 |
than 2 or more volumes shoved into one group. You would probably be |
15 |
better off adding both disks to ove LVM VG and telling the system to do |
16 |
simple striping or mirroring. |
17 |
|
18 |
I always got the feeling that such RAID facilities were little more than |
19 |
a way to get Windows users to have one big disk instead of two smaller |
20 |
ones using the simplest eans possible. |
21 |
|
22 |
Proper hardware RAID using a proper RAID adapter is a whole different |
23 |
story. Those tend to have proper firmware and control the disks |
24 |
properly. They do the optimizations you expect, they will reorder writes |
25 |
and do something sane with reads, and are worth the money is you need |
26 |
RAID. However, most still only do a subset of desirable RAID, usually 0, |
27 |
1 and 5. |
28 |
|
29 |
Linux software RAID, done in-kernel, is an amazing compromise. It's |
30 |
almost as efficient as good hardware RAID, and covers all bases and RAID |
31 |
types. Yes, it is somewhat more complex and you do have to use |
32 |
management utilities when working with it. It's similarly complex to |
33 |
administrating say LVM (not the same, just same order of magnitude) and |
34 |
comes with the benefit of NOT costing half a server motherboard :-) |
35 |
|
36 |
|
37 |
> |
38 |
> On 6 July 2015 at 16:01, Peter Humphrey <peter@××××××××××××.uk |
39 |
> <mailto:peter@××××××××××××.uk>> wrote: |
40 |
> |
41 |
> On Monday 06 July 2015 10:19:36 Alex Thorne wrote: |
42 |
> > I guess I did mean mdraid, but would you mind explaining the difference |
43 |
> > (I've never used raid so don't know much about this)? Is dmraid deprecated |
44 |
> > in favour of mdadm? |
45 |
> |
46 |
> Dmraid is the fake RAID that's included on most motherboards these |
47 |
> days; it's |
48 |
> meant for use with Windows and is enabled (or not) in the BIOS. |
49 |
> There are |
50 |
> Linux drivers, but we're always advised to use mdraid instead. |
51 |
> Mdraid is all |
52 |
> in software spread over the kernel, udev and user space*; it's not |
53 |
> influenced |
54 |
> at all by Windows as far as I know. Mdadm is the user-space |
55 |
> administration |
56 |
> program that comes with mdraid. |
57 |
> |
58 |
> Mdadm creates /dev/mdX from one or more /dev/sdX or similar - e.g. |
59 |
> my /dev/md1 |
60 |
> is built on /dev/sd[ab]1; /dev/md5 is on /dev/sd[ab]5 and /dev/md7 is on |
61 |
> /dev/sd[ab]7. That last one also has LVM on it with a dozen or more |
62 |
> logical |
63 |
> volumes for segments of my overall file system. |
64 |
> |
65 |
> If you want to play with mdraid, the old Gentoo guide is succinct |
66 |
> but useful: |
67 |
> |
68 |
> http://wwwold.gentoo.org/doc/en/gentoo-x86+raid+lvm2-quickinstall.xml |
69 |
> |
70 |
> Well, it was, but suddenly it isn't there - even Google's search |
71 |
> results end |
72 |
> up in an empty page. |
73 |
> |
74 |
> Ah, I've found the new version at https://wiki.gentoo.org/wiki/LVM . |
75 |
> It must |
76 |
> be very new - would you like to test it? :-) |
77 |
> |
78 |
> * Yes, I know that udev runs in user space (="User Device" |
79 |
> management) but I |
80 |
> thought it was worth mentioning separately. |
81 |
> |
82 |
> -- |
83 |
> Rgds |
84 |
> Peter |
85 |
> |
86 |
> |
87 |
> |
88 |
|
89 |
|
90 |
-- |
91 |
Alan McKinnon |
92 |
alan.mckinnon@×××××.com |